Explore the sounds of the early web with The MIDI Archive

Searching the future for what resides in the past

Archives seem to have come into a kind of fashion as of late, and I wonder if that has something to do with the socioeconomic legacies of the 2008 economic crash, or the end of the end of history. There's something knowable, certain, and grounded about the idea of archives, especially at a time in which precarity is the norm (interms of both economic prospects and more existentially). The training procedures of AI foundation models like GPT love archives too, but for their accessibility, volume, and givenness, and instead call them training sets.

The MIDI files collected here were once very new, not very long ago, and I've been sifting through them in search of the feeling of technological transformation, how they unite the alluring possibilties of new aesthetic experiences with the technics of producing and distributing this particular format of media.

Beyond the veil of nostalgia, a home to which one cannot return, this project presents a naive machine learning model alongside an archive. The two have developed alongside each other, and shape each other sensibilites, not only in terms of the quality of their presentation, but also in their shared energy. There exist numinous transformational qualities we ascribe to both the archive and machine learning, in the shadow of which we harbor hopes and fears that they may change something about ourselves.

Technical notes

The neural net model used here is pretty naive by design, and you can check out and run the source code yourself right from your browser. My intent is to have the model be able to express something whimsical, ineffable, and trues about the underlying archive, and not to be used seriously as a tool for music creation. Furthermore, I planned for this model to be portable and cheap to run. In its current implementation, it exists as a ~800k parameter decoder-only transformer model that runs daily from an AWS Lambda Function around noon (GMT) to produce a piece of music around three minutes in length.

The music produced by the model is available for your listening pleasure, along with the ~3000 MIDI files collected in this archive to train the model. This project also owes a debt to BitMIDI, which also developed the MIDI player I've implemented on this site, and allows visitors to listen to the 50,000 MIDI files archived from GeoCities.

A brief history [1980 — 2000]

Before MP3s came to dominate how people would listen to music on the internet, the sounds of the early web (and even BBS and Usenet before the world wide web) were predominantly expressed via MIDI. Its tiny file-size was accomodated by bandwidth limitations of the 1980s and 90s, web-native support for the format came early from browsers like Internet Explorer and Netscape Navigator.

There are currently 58862 MIDI files in the collection!

  • In addition to browsing the archive below, click here to listen to music produced by the neural net model. This music is generated daily at noon (GMT), so check back again later!
  • Table of contents for the archive:

    Bach MIDI Sequences by John Sankey

    John Sankey - Harpsichordist to the Internet

    MAHAMIDI --- Midi Files of John Mclaughlin

    The Play List For The Vietnam Veterans Home Page

    Textes et Musiques du Moyen Age

    The Israeli Jewish Yiddish Hebrew Folk Cultural Music Midi Free Library

    Laura's Midi Heaven

    Elephant Talk - The Official King Crimson Website

    This page is dedicated to MIDI and .WAV files of Robert Fripp, King Crimson and other related interests.

    Bot Productions' MIDI Collection

    Computer Generated MIDI MUSIC by Colin M. Johnson

    Christian MIDI Archive

    The Internet Renaissance Band

    Early music midi files by Curtis Clark

    Frank Lennon's IRISH MIDI FILES

    Prairie Frontier